Developing our preservative packaging through an iterative design
Compared to other food groups, fruits and vegetables have a relatively short shelf life due to their high moisture content (1). Fungal and bacterial pathogens are a major cause of fruit and vegetable spoilage, and therefore contribute to significant waste and income loss by producers and consumers. Post-harvest treatments such as the use of agrochemicals and thermal processing aim to reduce spoilage; however, between 25-40% of fruits and vegetables are still thrown out prior to consumption (1). Additionally, these preservation methods can lead to increased fungal resistance against chemical fungicides and decreased quality of produce, highlighting the need for a better, more environmentally-friendly solution.
Antimicrobial peptides (AMPs) are relatively small amino acid chains which can effectively target these pathogens and extend the shelf life of produce (2). AMPs can be obtained directly from animal, plant, or bacterial sources. However, doing so is costly and inefficient on an industrial scale.
But luckily, synthetic biology can allow for mass recombinant production of AMPs.
When deciding how to make our fruit package preservative, we knew we needed an AMP that was food-safe, targeted against a broad spectrum of produce-specific pathogens, and able to withstand various transportation conditions. We initially explored the option of expressing recombinant lysozyme - a food-safe, antimicrobial enzyme - in Escherichia coli (3). However, after meeting with our advisors and receiving feedback from judges at the Mindfuel Tech Futures Challenge competition, we realized that lysozyme is toxic to E. coli and would require an inhibitor to be co-expressed. Given our limited iGEM lab time, our team decided that we needed an alternative AMP that was non toxic to E. coli to demonstrate recombinant AMP expression for our application as a preservative.
Our search narrowed in on nisin.
Using our involutional neural network (read more about our INN here), nisin came up as one of our top hits. Nisin is a food-safe (FDA-approved) antimicrobial peptide which is commonly used as a biopreservative in dairy and poultry products (4). Derived from the Lactococcus lactis strain, nisin is relatively short, with only 34 amino acids in length. It consists of 5 ring structures (A through E) with several unusual post-translationally modified amino acids. Among its different variants, literature suggests that nisin Q (NisQ) has high antimicrobial activity and can inhibit oxidation better than its nisin A and nisin Z derivatives (5).
Nisin targets Gram-positive bacteria by binding to lipid II on the pathogen’s membrane, creating a pore in the cytoplasmic membrane and inhibiting peptidoglycan synthesis, thereby causing cell death (7). Research shows that nisin-producing bacteria can inhibit a small variety of fungi strains as well; however, there is limited literature on testing nisin against common fruit fungi (8). Unlike other AMPs, nisin is non-toxic to Gram-negative bacteria, meaning that we can achieve successful recombinant expression in E. coli without an inhibitory protein.
Figure 1. Nisin’s structure and mechanism of action. Adapted from Weidemann et al.
Importantly, nisin is stable at a broad range of pH and temperatures. Nisin’s optimal pH stability is between 2 and 7 but it can maintain its antibacterial activity up to a pH of 12, which is compatible with our bacterial cellulose NaOH purification steps (5). It can also retain its antimicrobial activity at temperatures of 120oC. This means that nisin can survive autoclaving, which was a critical consideration in designing how our AMP would be released into our BC packaging material during co-culture growth. These qualities make nisin an ideal AMP candidate for Cellucoat. Read more about our co-culture design and learn about how we integrated nisin into the fibres of our BC packaging.
Upon a literature review, we discovered that nisin Q has been expressed recombinantly in E. coli before, so we used this information to design our nisin-based Biobricks (5). We initially designed 2 unique composite parts. GB1 (BBa_K4437001) includes all necessary regulatory elements, a 6XHis-tag for protein purification, enterokinase cut sites for his-tag cleavage, and a TEV protease site (this was to ensure consistency with our GB2), the antimicrobial portion of nisin Q, and a T7 terminator. GB2 (BBa_K4437002) is identical to GB1, with one crucial difference - we included N-utilizing substance A (NusA), a functional sequence to help increase soluble nisin peptide in E. coli, because nisin is so small and may be difficult to express (5). We learned about the difficulty of bacterial secretion systems from our meeting with Dr. Burkinshaw, a biology and biochemistry professor at the University of Calgary, which informed our desision to include NusA. For ease of expression in various vectors, we made sure that our restriction enzyme cut sites were not included in our G-blocks, and instead are flanking the primer ends as overhangs. Read more about our part design for NisQ.
Figure 2. Our genetic constructs designed to express NisQ
Although previous iGEM teams have created nisin DNA constructs, we are the first iGEM team to design Biobricks for NisQ. In 2014, the iGEM14_Groningen team designed a Biobrick for nisin A (NisA) and were able to transform E. coli BL21 cells. Other teams, such as the 2021 IISER_Kolkata have expressed NisA using a natural producer of nisin, the L. lactis bacteria strain. Our goal with expressing recombinant NisQ is to demonstrate that production in E. coli can allow for mass production (which is why we chose not to use the L. lactis strain), and because NisQ has greater antimicrobial activity compared to NisA.
Our 3rd composite part GST-His-NusA-NisQ-His, derived from GB2, was not created until later into the iGEM season. Read more about BBa_K4437003 here.
We successfully PCR amplified both our NisQ DNA insert, known as GB1 (BBa_K4437001) and NusA+NisQ DNA insert, known as GB2 (BBa_K4437002) with little to no troubleshooting. This allowed us to quickly move onto digesting both the inserts and a pSB1A3 backbone for subsequent ligation and transformation into chemically competent Top 10 E. coli cells.
However, the following cloning steps proved to be more difficult than anticipated. After cloning several attempts due to failed transformations and unsuccessful plasmid minipreps we were finally able to run a PCR amplification with construct specific primers and gel that confirmed our expected band sizes within pSB1A3 (see Figure 3). We then confirmed that we did not accidentally introduce point mutations or indels into our sequence, using DNA sequencing. We demonstrated that we were able to successfully clone BBa_K4437001 into an expression vector for subsequent protein expression.
Figure 3. PCR amplification of pSB1A3 plasmid containing GB1, using T7 promoter and terminator specific primers (GB1 = 393 bp), indicating successful cloning.
To produce protein from our successfully cloned construct, we transformed chemically competent E. coli BL21 with our isolated pSB1A3 plasmid containing GB1. Testing both Overnight ExpressTM autoinduction media and two types of ZYP media (ZYP-0.8G and ZYP-5052) which varied in glucose percentages, we tried to find optimal conditions that would allow for protein expression. However, our SDS-PAGE gel yielded no clear bands at the expected 7.0 kDa range for GB1 construct containing NisQ. We suspected that because our protein was so small it may be hard to visualize, so we used an anti-His primary antibody and HRP conjugated secondary antibody for Western blotting. Unfortunately, no bands for NisQ were observed on our Western blot either.
After consulting our wetlab advisors, we decided to go back and attempt re-cloning our GB2 construct into a different plasmid, called the Xpress expression vector (BBa_K3945014), characterized by the 2021 iGEM Calgary team. Xpress contains a GST solubility tag which can help fold our nisin protein better. We successfully digested and ligated our GB2 into the Xpress vector using HindIII and BamHI cut sites, and cloned our plasmid into chemically competent E. coli BL21. Fusing GST to NusA within our GB2 construct formed our third composite part in our NisQ part collection(BBa_K4437003), and allowed us to increase our expected protein size from 7kDa to 91kDa.
Figure 4.1 SDS-PAGE analysis of GST-NusA-NisQ (BBa_K4437003) samples from BL21 (DE3) E. coli strain autoinduced, using a Coomassie-blue stain. Large bands in lanes 2, 4, 5, 6, and 9 at 91kDa correspond to our expected protein size.
Figure 4.2 His-tag purified SDS-PAGE of GST-NusA-NisQ (BBa_K4437003) samples samples from BL21 (DE3) E. coli strain autoinduced, using a Coomassie-blue stain. Bands in lanes 3, 4, and 5 at 91kDa correspond to our expected protein size. Samples labelled "W-1" indicate wash 1, samples labelled "E-1" indicate elution 1.
A western blot of the whole cell lysate was done in tandem with the SDS-PAGE of the whole cell lysate to solidify that the protein of interest was present in case the SDS-PAGE analysis of the batch NI-NTA purification of the his-tagged GST-NusA-NisQ did not show detectable results.
Below shows the correct protein size visible between 110 & 80 kDa (~ 91 kDa) for the three instances of GST-NusA-NisQ. The band sizes shown below mirrors the band sizes of the SDS-PAGE gel of the NI-NTA purification of the his-tagged GST-NusA-NisQ samples, which solidified that protein of interest was expressed.
Figure 5. Western blot of the whole cell lysate of GST+NusA+NisQ auto-induced in overnight express. A his-tagged positive control was also included. The protein ladder used was the Novex sharp pre-stained protein standard. The antibodies used were Mouse Anti-HIS-tag mAb (Abcam) for the primary antibody and Goat Anti-Mouse:HRP (Abcam) for the secondary antibody.
Our plan moving forwards is to purify our NisQ protein from NusA and GST using TEV protease and enterokinase, allowing for Ni-NTA purification with our 6XHis-tag. Once purified, we can characterise our protein's antimicrobial properties using a Kirby-Bauer disc diffusion test against B. subtilis. We also intend to use our transformed E. coli containing our NisQ sequence into a co-culture with K. xylinus to integrate nisin evenly throughout our BC packaging. Read more about our co-culture design here.
After purchasing nisin solution, we began running a series of characterization tests against various fungi strains collected from either rotting fruit or samples provided to us by Francene Cusack and Dr. Heather Addy from the Biological Sciences Department at the University of Calgary. We attempted to run a Kirby-Bauer disc diffusion test to determine the effectiveness of nisin against our known fungi strains along with the unknown fungi swabbed off of rotting fruit collected from a local grocery store.
Figure 6. Rotting fruit collected from our local grocery store. Fruit fungi and bacteria swabbed and plated.
However, we quickly realized that our lab was not equipped to handle fungi safely, so we decided to send our fungi samples and nisin solution to the U of Alberta iGEM 2022 team, as their lab had the proper equipment and supervisor expertise to safely handle BSL2 fungi. To read more about our partnership with the U of Alberta iGEM 2022 team, click here.
Luckily, nisin has a wide range of targets beyond fungi - specifically gram positive bacteria - allowing us to test its effectiveness against more than one foodborne pathogen (9). We met with Josh McGinnis, who is a bioprospector and the CEO of EveryMan Bio, to help us determine which potential bacteria species were present on our swabbed grocery store fruits and whether nisin could target the species. Josh identified Bacillus bacteria on our rotting stone fruits, which informed our decision to test nisin’s effectiveness against Bacillus subtilis. B. subtilis, while not harmful to humans, is a Gram-positive bacteria that nisin can effectively target and thus demonstrate Cellucoat’s antimicrobial potential against other Gram-positive bacteria (10).
We ran a series of Kirby-Bauer disc diffusion tests comparing nisin to other known antimicrobial substances found in literature, including phenol and thymol. To simulate Cellucoat’s integration of nisin into bacterial cellulose (BC), we used our lab-grown BC to create paper discs for our Kirby Bauer tests.
Figure 7. Initial average zone of inhibition sizes for Kirby Bauer disc diffusion test results against B. subtilis. Diameter measurements (n=3) were taken 24 hours after plating using a ruler. Error bars represent standard deviation. Asterisks represent a significant difference between treatments (p<0.05).
Our results from our initial round of testing indicated that nisin was effective against B. subtilis but not to the same extent as phenol and thymol. Statistical analysis using a two sample t-test revealed that there was a significant difference in average zone inhibition sizes between groups. Notability, there was also a significant difference in the average zone inhibition sizes between each nisin group. After consulting one of our mentors, Andrew Symes, we knew that the difference was either attributed to a difference in biology or measurement but we could not confidently say which. Because we had taken our measurements using a ruler and could only approximate the zone inhibition sizes to the nearest tenth of a cm, this presented a problem. Our methodology was not very precise and we needed a standardized method to collect data.
While KB-Perry cannot sing “Firework”, it does have the capacity to ensure consistent measurements for our Kirby Bauer disc diffusion tests. We took measurements of the dimensions of our petri dishes used to plate our bacteria, along with the distance between our phone camera and the plate. Using this information, we created a model on Fusion360 and then 3D printed an adjustable phone stand. By utilizing KB-Perry, diameter measurements could then be taken using Fiji ImageJ software, which could capture the zone of inhibition in pixels (1mm = 3.3 pixels) - proving to be far more precise than a ruler. Read more about our measurement system here.
Now that we had a more consistent measurement system, all of our subsequent Kirby Bauer experiments were performed using KB-Perry.
If Cellucoat is going to be integrated into the food packaging industry, it must retain its preservative properties over time. Microbes that are responsible for spoilage may come into contact with fruit surfaces at any time post-harvest - even before packaging (11). In our meetings with Chris Clark from Star Produce, and Chris Messent from Consolidated Fruit Packers Inc., we learned about the importance of transport conditions in our packaging design. To test whether nisin can target and inhibit the growth of Gram-positive bacteria (even when the bacteria had been growing at its optimal conditions for several hours), we ran a series of Kirby-Bauer disc diffusion tests over time. After plating B. subtilis (time point = 0 hrs) we placed our lab-grown BC paper discs soaked in either nisin, phenol, thymol, or water, in a separate quadrant of the plate. All four treatments were applied at 2 hour intervals for a total of 6 hours.
Figure 8. Average zone of inhibition sizes for Kirby Bauer disc diffusion test results against B. subtilis over 6 hours. Diameter measurements (n=3) were taken 24 hours after plating using KB-Perry. Error bars represent SEM.
Although nisin was consistently less effective against B. subtilis compared to thymol and phenol, we still demonstrated nisin’s success as an antimicrobial. The goal of this set of experiments was to simulate a scenario where bacteria had already begun growing on the surface of fruit, and test whether Cellucoat could effectively kill existing colonies and prevent further growth. Our results showed that nisin remained active against B. subtilis after 2, 4, and 6 hours of optimal bacterial growth conditions. While this may seem like a relatively small time frame, it is reasonable to argue that the growing conditions for this experiment will not be comparable for fruit post-harvest and during transport. Trucks maintain their temperatures at a chilly 0oC to 12oC, unlike our incubator which is kept at 37oC (12). Therefore, the expected window of time for nisin’s activity for fruit in transport will likely be much longer than in a lab.
To further investigate Cellucoat’s capacity for preservation, we conducted a spot test on grapes by immobilizing nisin onto our lab-grown, co-cultured BC. This experiment was informed by the feedback we recieved at our Faculty Talk. We created a timelapse video that spanned over the course of 12 days and observed whether our packaging would prevent rotting and spoilage. After 12 days, we removed our BC samples to reveal that they had noticeably prevented deterioration compared to the uncovered surface of the grape. However, there was no noticeable difference in the BC + nisin spot and the BC alone spot.
To develop an effective antimicrobial material and reduce the risk of spoilage, we needed a means to identify the concentration of nisin necessary to lyse microbes. Our solution? A minimum inhibitory concentration (MIC) test, a widely used method which is currently the best known procedure to determine the minimum amount of antimicrobial substance required to prevent growth of bacteria. Adding varying concentrations of antimicrobial substance and taking OD measurements after an incubation time for a sample of bacteria culture can determine how effective said substance is at preventing growth. The MIC value is found to be the concentration of antibacterial substance that produces an OD600 value that is less than or equal to the measurement of a sterile sample (13).
An MIC test involves the use of serial dilutions to add increasingly smaller amounts of antibiotic to the wells of a 96-well plate. An equivalent amount of bacterial culture is then added and incubated. Using a 96-well plate reader, the OD600 of these wells is then measured to quantify the amount of growth inhibited by the chosen antimicrobial agent.
In our experiments, we tested the effectiveness of nisin against B. subtilis. As a control, we ran parallel tests with ampicillin against E. coli. First, we selected a suitable concentration range of the two test solutions: we settled on a range of 0.64000 μg/mL to 0.00125μg/mL for ampicillin, and a range of 2,000-4,000 IU/mL to 0.98-1.95 IU/mL for nisin; using a two-fold serial dilution (our stock nisin had a listed concentration of 20,000-40,000 IU/mL). Our results indicated that, for ampicillin, a concentration of 0.32 μg/mL yielded an OD600 less than that of our sterile well (OD600 = 0.0919 ± 0.0192), giving a absorbance value of OD600 = 0.0517 ± 0.0157. For nisin, the MIC value was found to be a 1/320 dilution, with a concentration of 62.5-125 IU/mL. This gave an OD600 of 0.0177 ± 0.0011. To perform a better comparison we converted the concentration of nisin in IU/mL to μg/mL using the conversion 1 IU of nisin is equivalent to 0.025 μg of nisin (14). We therefore get a MIC value for nisin to be 1.56-3.13 μg/mL. For comparison, uninhibited growth had an OD600 of 0.5962 ± 0.0320.
Figure 10. MIC results for nisin against B. subtilis
Figure 11. MIC results for ampicillin against E. coli
In comparing the MIC of ampicillin (0.32 μg/mL) and nisin (1.56-3.13 μg/mL) we see that nisin has a comparable inhibition of B. subtilis growth with a concentration that is 4.9-9.8 times greater than ampicillin.
We know that nisin isn’t effective against all bacteria and fungi, which means fruits can still be vulnerable to other pathogens. Additionally, while nisin currently serves as a great option to functionalize cellulose for antimicrobial properties, we know that AMP discovery is ongoing. There may be better AMPs aside from nisin that would better serve our product’s purpose, so we wanted to ensure that future iterations of Cellucoat can use the best AMPs.
With these considerations in mind, we designed Cellucoat to be both customizable and modular.
But with over 880 different AMPs in literature, we realized we needed an efficient method to sift through every possible peptide. We programmed a novel, involutional neural network (or INN) that is able to identify key patterns present in AMPS based on their DNA sequence. Our INN currently performs as accurately as a conventional CNN, and we are conducting more tuning to try and improve it further. Read more about our INN here.
To create a modular expression system, we have also decided to take advantage of Golden Gate Assembly - where BsaI restriction enzymes are able to cut downstream of their recognition site in a given DNA sequence. This allows for many antimicrobial peptide DNA sequences to be compatible with both various signal peptide sequences, and our plasmid backbone. Read more about our use of Golden Gate here.